Cocktail: Learn a Better Neural Network Controller from Multiple Experts via Adaptive Mixing and Robust Distillation

TL;DR

COCKTAIL is a framework that learns a robust, efficient neural network controller by adaptively combining multiple experts and distilling their knowledge, improving safety and verifiability in cyber-physical systems.

Contribution

It introduces a novel adaptive mixing and distillation approach to synthesize neural controllers from multiple experts, enhancing robustness and efficiency.

Findings

Significant improvement in control robustness against adversarial attacks.

Enhanced energy efficiency of the learned controllers.

Reduced verification computation time compared to baselines.

Abstract

Neural networks are being increasingly applied to control and decision-making for learning-enabled cyber-physical systems (LE-CPSs). They have shown promising performance without requiring the development of complex physical models; however, their adoption is significantly hindered by the concerns on their safety, robustness, and efficiency. In this work, we propose COCKTAIL, a novel design framework that automatically learns a neural network-based controller from multiple existing control methods (experts) that could be either model-based or neural network-based. In particular, COCKTAIL first performs reinforcement learning to learn an optimal system-level adaptive mixing strategy that incorporates the underlying experts with dynamically-assigned weights and then conducts a teacher-student distillation with probabilistic adversarial training and regularization to synthesize a student neural network controller with improved control robustness (measured by a safe control rate metric with respect to adversarial attacks or measurement noises), control energy efficiency, and verifiability (measured by the computation time for verification). Experiments on three non-linear systems demonstrate significant advantages of our approach on these properties over various baseline methods.